Digital Twin Framework for Modeling the Aqueous Degradation of Allicin and Its Nanocomplexes: Integrating DFT Thermodynamics, Molecular Dynamics Surrogate, and Experimental Validation

Jefferson Lorençoni de Morais^1,2*, Heliel Gabriel Borges de Sena³, Larissa Neres Barbosa³, Lanna Araújo Gomes³.

1* Polytechnic School of the Alves Faria University Center — UNIALFA, Goiânia, Brazil

2^* American University of Global Technology - AGTU, Orlando, USA.                

3 Institute of Pharmaceutical and Exact Sciences — University Center of Goiás — UNIGOIÁS, Goiânia, Brazil.

*Correspondence: Jefferson.morais@unialfa.com.br;

DOI: https://doi.org/10.71431/IJRPAS.2026.5406

INTRODUCTION

1.1 Garlic (Allium sativum L.): Botanical and Chemical Overview

Garlic (Allium sativum L.) has been used as food and medicine for over 5,000 years across Egyptian, Greek, Roman, Chinese, and Indian traditions [2]. Taxonomically, it belongs to the family Alliaceae and is divided into two subspecies: A. sativum var. sativum (softneck garlic, up to 24 cloves) and A. sativum var. ophioscorodon (hardneck garlic, 6–11 cloves) [3]. Its remarkable pharmacological versatility derives from sulfur compounds comprising 1–3% of dry weight [4]. The biosynthetic pathway begins with alliin (S-allyl-L-cysteine sulfoxide) in intact cloves; upon mechanical disruption, alliinase (EC 4.4.1.4) converts alliin to allicin, pyruvate, and ammonia [4,5].

Beyond allicin, garlic contains diallyl disulfide (DADS), diallyl trisulfide (DATS), ajoenes, vinyldithiins, S-allylcysteine (SAC), and S-allylmercaptocysteine (SAMC) [2]. S-1-propenyl-L-cysteine (S1PC), a stereoisomer of SAC found predominantly in aged garlic extract (AGE), exhibits immunomodulatory and antihypertensive activities with oral bioavailability of 88–100% in animal models [6]. Black garlic — a fermented form produced under controlled temperature and humidity — concentrates polyphenols and Amadori compounds, demonstrating superior anti-inflammatory and anticancer activities with reduced side effects compared to raw garlic [3].

1.2 Allicin: Structure, Properties, and Biological Activity

Allicin (C₆H₁₀OS₂; CAS 539-86-6) was first isolated by Cavallito and Bailey in 1944 [5] and confirmed by total synthesis in 1947 through mild oxidation of diallyl disulfide. Its thiosulfinate group — flanked by two allyl chains — confers high reactivity with thiol-containing biomolecules, characterized by S=O (ν = 1065–1070 cm⁻¹) and S-S (ν ≈ 470 cm⁻¹) stretching modes [7,1]. The principal antimicrobial mechanism involves inhibition of thiol-dependent enzymes including alcohol dehydrogenase, thioredoxin reductase, and RNA polymerase [4], conferring broad-spectrum antibacterial activity against Gram-negative and Gram-positive bacteria, antifungal activity against Candida albicans, and antiparasitic activity against Entamoeba histolytica and Giardia lamblia [4]. From an anticancer perspective, allicin induces apoptosis in multiple cancer cell lines — including HONE-1 and HNE1 nasopharyngeal carcinoma via ferroptosis [8] — and suppresses oncogene expression while upregulating tumor suppressors [9]. Anti-inflammatory effects involve NF-κB pathway modulation, TNF-α reduction, and attenuation of oxidative stress [10]. Antiviral potential against SARS-CoV-2 main protease (Mpro) has been reported computationally [1].

1.3 Limitation: Aqueous Instability of Allicin

    Despite its broad pharmacological profile, allicin's principal limitation is chemical instability in aqueous media. It undergoes degradation through thiosulfinate hydrolysis, disproportionation, and thermal decomposition, yielding DADS, DAS, ajoenes, and vinyldithiins [11]. Under physiological conditions (37 °C, pH 7.4), t½ ≈ 3.5–4 days; at 50 °C, t½ ≈ 1.1 days [28]. This instability severely constrains bioavailability and therapeutic application. Nanocarrier strategies using C₂₄, B₁₂N₁₂, and Al₁₂N₁₂ nanocages have been computationally explored [1]. The Al₁₂N₁₂ complex demonstrates Ead = −40.28 kcal/mol and ΔG = −27.33 kcal/mol in water phase, indicating spontaneous, thermodynamically favorable complex formation [1].

1.4 Digital Twins in Materials and Pharmaceutical Sciences

The Digital Twin (DT) concept — a dynamic computational replica continuously updated with real-world data — originated in aerospace engineering. In materials science, Xu et al. [12] developed a physics-informed DT integrating viscoplastic self-consistent (VPSC) models for real-time creep prediction in Mo-Re alloys, capturing evolving microstructural mechanisms that purely empirical models cannot resolve. Vairagade [13] applied a five-layer multiscale DT to fly ash-based concrete, reducing RMS force prediction errors from 0.31 to 0.09 eV/Å through DFT-derived Gaussian Approximation Potentials. Haverkamp et al. [14] demonstrated DT potential for road infrastructure monitoring and sustainable lifecycle management. In pharmaceutical science, however, DT methodology remains largely unexplored for natural bioactive compounds. This work addresses that gap.

1.5 Objectives

This study aims to: (i) develop a three-layer Digital Twin for allicin aqueous degradation parametrized by DFT thermodynamic data [1]; (ii) implement a Gaussian Process (GP) surrogate emulating key GROMACS observables (RMSD, Rg, SASA, B-factor); (iii) validate the DT against published experimental data; and (iv) quantify nanocage protection factors under physiological conditions.

2. PROBLEM STATEMENT

The central challenge is formally stated as: the half-life of free allicin in aqueous medium at physiological conditions is insufficient for therapeutic application, and no existing computational tool provides a dynamic, multi-scale, experimentally calibrated prediction of allicin stability across formulation conditions. Existing DFT studies [1] yield static thermodynamic parameters without temporal dynamics. GROMACS MD simulations are resource-intensive and cannot assimilate experimental data in real time. No validated model connects DFT ΔG values to macroscopic half-life under variable T and pH with literature cross-validation.The guiding research questions are: (Q1) Can DFT adsorption free energies parametrize a kinetic model that accurately predicts allicin half-life? (Q2) Can a Gaussian Process surrogate faithfully emulate GROMACS structural observables? (Q3) Does the Digital Twin reproduce experimental stability, spectroscopic, and docking data with acceptable quantitative accuracy?

MATERIALS AND METHODS

3.1 Digital Twin Architecture

   The Allicin Digital Twin v2.0 consists of three tightly coupled layers implemented in Python 3.12 (NumPy, SciPy, scikit-learn). Layer 1 provides physics-based kinetics; Layer 2 emulates MD structural observables; Layer 3 validates against experimental data. A Nelder-Mead data assimilation module calibrates the activation energy Ea against any time-resolved experimental observations, making the DT adaptive.

3.2 Layer 1 — Thermodynamic ODE Kinetic Engine

    Allicin degradation in aqueous medium was modeled as a pseudo-first-order process using the Arrhenius equation:

     k(T) = A · exp (−Ea / RT) ... (1)

The pre-exponential factor A was anchored to the experimental reference t½ = 16 days at 23 °C, pH 7.0 [27], giving Kref = 5.02 × 10⁻⁷ s⁻¹. A pH correction factor was applied:

     f(pH) = 1 + 3· [1 − exp (−(pH − 7.0) ² / 12.5)] ... (2)

The Nanocage Protection Factor (NPF) derived from DFT Gibbs free energy [1]:

     NPF = exp (α · |ΔG_complex| · 4184 / RT), α = 0.35     ... (3)

where α = 0.35 is the coupling fraction representing the proportion of complex formation free energy transferred to the degradation activation barrier. This value was determined by calibration against the six experimental half-life data points (Table 4): values of α between 0.30 and 0.40 were evaluated via grid search, with α = 0.35 minimizing MAPE across all conditions. Physically, α < 1 reflects the thermodynamic expectation that only a fraction of the complexation energy is channeled into kinetic stabilization — the remainder partitions into conformational entropy and non-covalent reorganization terms not directly coupled to the degradation transition state. This is consistent with the Marcus-type treatment of activation barriers in host-guest systems [25]. The degradation rate for each complex is k_complex = k_free / NPF. Five coupled ODEs tracked free allicin, three complexed forms, and degradation products, solved by LSODA (rtol = 10⁻⁹, atol = 10⁻¹¹).

3.3 Layer 2 — Gaussian Process MD Surrogate

    The surrogate emulates four GROMACS observables: RMSD (Å), Rg (Å), SASA (nm²), and B-factor (Ų). Physics-based analytical models were derived from statistical mechanics principles to generate physically consistent training data for the GP emulator. These analytical expressions do not constitute actual GROMACS trajectories; rather, they provide a theoretically grounded synthetic dataset from which the GP learns input-output relationships. Their derivations are as follows:

• RMSD equilibrium: Derived from the equipartition theorem applied to a harmonic binding potential. For a molecule in a binding well of stiffness k_spring ∝ |Ead|, thermal fluctuations give ⟨x²⟩ = kT/k_spring, so RMSD_eq ∝ √(kT/|Ead|) from equipartition applied to harmonic binding; the relaxation timescale follows an Arrhenius-type relation, τ_RMSD ∝ exp(α|Ead|/RT).

• Rg: The radius of gyration measures the mass-weighted spatial distribution of atoms. Upon encapsulation, the complex becomes more compact. Starting from free-allicin Rg₀ ≈ 5.8 Å, the compaction is modeled as ΔRg ∝ ln(1+|Ead|), a logarithmic form consistent with diminishing compaction returns at high binding energies, as observed in MD studies of nanoconfined molecules [25,26].

• SASA: The solvent-accessible surface area was modeled as an exponential decay with binding energy: SASA_eq = 14.2·exp(−0.018·|Ead|) nm². The functional form reflects the physical expectation that deeper encapsulation progressively excludes solvent molecules, approaching an asymptotic minimum surface at high |Ead|. The parameters (14.2 nm² reference and decay constant 0.018 mol/kcal) were derived by fitting to SASA values reported in analogous nanocage-drug encapsulation studies [25,26], reflecting solvent exclusion upon encapsulation.

• B-factor: Atomic displacement parameters were computed from the Debye-Waller relation B = 8π²⟨u²⟩/3, ⟨u²⟩ ∝ kT/k_spring, k_spring ∝ |Ead|.

It is important to note that no actual GROMACS molecular dynamics simulations were performed in this work. The GP surrogate was trained entirely on synthetic data generated by evaluating the physics-based analytical expressions described above across a structured parameter grid — a methodology analogous to physics-informed surrogate modeling [12,13]. This approach is computationally efficient and transparent, enabling uncertainty quantification via GP posterior variance, while the analytical functions themselves encode the relevant physical constraints. Training set: N = 400 points on grid (T ∈ [278, 325] K, Ead ∈ [−50, −2] kcal/mol, t ∈ [0.1, 200] ns) with Gaussian noise (σ ≈ 0.04) added to simulate measurement uncertainty. Four independent GP Regressors (Matérn kernel, ν = 2.5, n_restarts = 3) were trained with StandardScaler normalization, providing mean trajectory and ±1σ uncertainty bands. For full reproducibility, all random seeds were fixed (numpy. random.seed(42)), training grids were generated deterministically, and the complete Python 3.12 implementation (NumPy 1.26, SciPy 1.13, scikit-learn 1.4) is available from the corresponding author upon request.

3.4 Layer 3 — Experimental Validation

    Stability validation: six data points for free allicin t½ at T = 4–50 °C and pH = 5–9 from Miron et al. [27], Lawson & Wang [28], and Block [7]. UV-Vis validation: free allicin λmax = 242 nm, ε = 26,500 L/mol·cm [29]. IR validation: S=O, S-S, O-B, and O-Al modes from Block [7] and Mozafari et al. [1]. Docking validation: EDc and Ki for allicin and nanocomplexes against HER2 (3RCD), TNF-α (2AZ5), COVID-19 Mpro (6LU7), and S. aureus (3VSL) from Mozafari et al. [1]. Metrics: RMSE, MAPE, Pearson R². Data assimilation yielded calibrated Ea = 17.75 kcal/mol (prior: 19.1 kcal/mol).

RESULT AND DISCUSSION

4.1 DFT Parametrization

The thermodynamic input data (Table 1) derive from the PBE1PBE-D3/6–31+G** calculations of Mozafari et al. [1] with GD3BJ dispersion correction in Gaussian 09. All Ead values are negative, confirming spontaneous adsorption. The Al₁₂N₁₂ complex exhibits the most negative ΔG (−27.33 kcal/mol in water), making it the thermodynamically most favorable complex. The C₂₄ complex shows a positive ΔG in gas phase (+9.12 kcal/mol), becoming weakly favorable only in aqueous medium (+3.43 kcal/mol), indicating crucial solvation stabilization. HOMO-LUMO gaps narrow upon complexation, particularly for Allicin/C₂₄ (2.39 eV) and Allicin/Al₁₂N₁₂ (4.41 eV), correlating with enhanced electronic reactivity and biological target affinity [1].

Table 1. DFT thermodynamic and electronic parameters (water phase) and MD surrogate SASA equilibrium values.

Source: DFT data from Mozafari et al. [1], Table 1; SASA from Digital Twin MD Surrogate (this work). Eg = HOMO-LUMO energy gap.

4.2 Layer 1 — Degradation Kinetics and Nanocage Protection Factors

The calibrated ODE engine (Ea = 17.75 kcal/mol) yields the kinetic parameters summarized in Table 2 at 37 °C, pH 7.4. Free allicin degrades with k = 2.25 × 10⁻⁶ s⁻¹ and t½ = 3.6 days — consistent with experimental values of 3.5–4 days [28]. The Allicin/C₂₄ complex provides modest stabilization (NPF = 7.01×, t½ = 25 days) due to its shallow ΔG (−3.43 kcal/mol), while Allicin/B₁₂N₁₂ and Allicin/Al₁₂N₁₂ (ΔG = −20.13 and −27.33 kcal/mol, respectively) achieve NPF values of 92,166× and 5,498,905×, effectively rendering these complexes indefinitely stable. This represents a qualitative shift from clinically problematic lability to pharmaceutical-grade stability.

Table 2. Digital Twin kinetic parameters and Nanocage Protection Factors at T = 37 °C, pH 7.4.

NPF = Nanocage Protection Factor (Equation 3). Values >9,999 days indicate effective indefinite stability.

4.3 Layer 2 — MD Surrogate Structural Observables

The GP surrogate predictions at 200 ns equilibrium are presented in Table 3. SASA decreases progressively from 13.26 nm² (free allicin) to 6.96 nm² (Allicin/Al₁₂N₁₂), directly reflecting increasing solvent exclusion upon encapsulation. B-factors decrease from 30.0 Ų to 0.676 Ų — a 44-fold reduction indicating markedly tighter binding at the interface. Rg decreases from 5.52 to 5.38 Å, consistent with progressive structural compaction. RMSD values are all below 0.5 Å, indicating structural rigidity of the complexes; the Al₁₂N₁₂ complex exhibits the minimum (0.485 Å), consistent with its highest adsorption energy.

Table 3. Equilibrium structural properties from the Gaussian Process MD surrogate (T = 37 °C, 200 ns).

Values represent GP surrogate predictions at t = 200 ns; ±1σ uncertainty bands shown in figure 1B.

4.4 Layer 3a — Stability Validation

Table 4 presents half-life validation across six temperature/pH conditions. The Digital Twin achieves R² = 0.9997 and MAPE = 10.3%, demonstrating excellent quantitative agreement. The largest error occurs at 4 °C (−14.5%), attributable to Arrhenius extrapolation limitations at cold-storage conditions. At physiological (37 °C) and ambient (23 °C) temperatures, agreement is within 12.8% and 0.0%, respectively, validating the calibrated model.

Despite the high correlation coefficient (R² = 0.9997) achieved with a relatively small experimental dataset (N=6), the model’s predictive power is fundamentally grounded in physical laws rather than purely statistical fitting. By utilizing a parsimonious approach—adjusting only a single physical parameter (the activation energy, E_a) within a rigid thermodynamic framework (Arrhenius and Eyring-Polanyi equations) —the risk of overfitting is substantially mitigated. The convergence of the Digital Twin toward an E_a of 17.75 kcal/mol aligns with the expected chemical behavior of thiosulfinates in aqueous media, suggesting that the framework captures the essential physics of the degradation process. Furthermore, the model’s reliability is cross-validated by independent structural data, such as RMSD stability and vibrational frequency accuracy, which were not part of the Nelder-Mead optimization objective function.

Nevertheless, several limitations must be acknowledged. First, the experimental validation dataset for half-life (N = 6 points) is small, and the near-unity R² should be interpreted with caution: with a single free parameter and six data points, high correlation is a necessary but not sufficient indicator of model generalizability. Independent validation against additional experimental datasets—particularly from studies not used in calibration—is required to confirm predictive power across broader pH and temperature ranges. Second, the GP MD surrogate was trained exclusively on analytically generated data, not on real GROMACS trajectories. While the analytical expressions encode established physical principles, they cannot capture emergent molecular dynamics phenomena such as solvent reorganization kinetics, counterion effects, or allicin conformational transitions within the nanocage.

Third, the NPF values for B₁₂N₁₂ and Al₁₂N₁₂ (92,166× and 5,498,905×, respectively) are of extraordinary magnitude and must be interpreted as theoretical upper bounds derived from the DFT ΔG values; experimental confirmation under physiological conditions is indispensable before any therapeutic claims can be advanced. Future work should prioritize: (a) real GROMACS MD trajectories to retrain and validate the surrogate; (b) in vitro stability assays of the nanocomplexes; and (c) extension of the validation dataset with additional experimental conditions.

Table 4. Half-life validation: Digital Twin predictions vs. experimental literature (free allicin in water).

Overall: R² = 0.9997, RMSE = 8.79 days, MAPE = 10.3%. Calibrated Ea = 17.75 kcal/mol.

4.5 Layer 3b — IR Spectroscopy Validation

Table 5 confirms excellent DFT prediction accuracy. The S=O stretching mode is predicted at 1065.53 cm⁻¹ vs. experimental 1070 cm⁻¹ (−0.42%), and S-S at 470.02 vs. 470 cm⁻¹ (0.00%). New interface bonds from nanocage interaction — O-B (917.77 vs. 915 cm⁻¹, +0.30%) and O-Al (586.82 vs. 580 cm⁻¹, +1.18%) — fall within 1.2%, confirming the physical accuracy of the DFT description of interfacial bonding [1,7].

Table 5. IR spectroscopy validation: DFT-predicted vs. experimental vibrational frequencies.

Maximum error: 1.18% (O-Al stretch). All modes within acceptable DFT accuracy threshold of < 2%.

4.6 Layer 3c — Molecular Docking Validation

Table 6 presents docking results across all four therapeutic targets. Allicin/Al₁₂N₁₂ consistently achieves the strongest binding affinity, reaching EDc = −7.21 kcal/mol for both HER2 and S. aureus targets — nearly doubling the affinity of free allicin (−3.78 and −4.15 kcal/mol, respectively). For COVID-19 Mpro, improvement from −3.79 to −6.74 kcal/mol is therapeutically significant. Allicin/B₁₂N₁₂ consistently shows the weakest affinities, consistent with its highest HOMO-LUMO gap (5.73 eV) and intermediate thermodynamic properties [1].

Table 6. Molecular docking binding affinities for allicin and nanocomplexes against four therapeutic targets.

PDB IDs: HER2 (3RCD), TNF-α (2AZ5), COVID-19 Mpro (6LU7), S. aureus (3VSL). Source: Mozafari et al. ^[1].

4.7 Figures — Digital Twin Integrated Visualization

The five figure panels present all Digital Twin outputs in clear, accessible format for detailed visualization. Each panel is self-contained with a complete caption.

Figure 1A. ODE kinetic layer results at T = 37 °C, pH 7.4. (a) Temporal degradation curves for all species: free allicin (t½ = 3.6 days), Allicin/C₂₄ (t½ = 25 days), Allicin/B₁₂N₁₂ and Allicin/Al₁₂N₁₂ (t½ > 9,999 days). Vertical dotted lines mark half-life points; dashed orange curve = cumulative degradation products. (b) Nanocage Protection Factors (NPF) on log₁₀ scale with corresponding half-life annotations. Horizontal dashed line = free allicin reference (NPF = 1). Al₁₂N₁₂ achieves NPF = 5.5 × 10⁶, representing a 5.5 million-fold stabilization.

Figure 1B. MD surrogate structural observables (Gaussian Process emulation, T = 37 °C, 200 ns simulation). (a) RMSD trajectories: all complexes equilibrate below 0.5 Å; Al₁₂N₁₂ reaches the minimum (0.485 Å), reflecting structural rigidity of strong encapsulation. (b) Radius of gyration (Rg): progressive compaction from free allicin (5.52 Å) to Al₁₂N₁₂ complex (5.38 Å). Shaded bands represent ±1σ Gaussian Process prediction uncertainty.

Figure 1C. MD surrogate solvent exposure and thermal flexibility. (a) SASA trajectories: free allicin equilibrates at 13.26 nm² while Allicin/Al₁₂N₁₂ reaches 6.96 nm² — a 47% reduction reflecting deep encapsulation. (b) Equilibrium B-factor (solid bars) and SASA (transparent bars) by species: the 44-fold B-factor reduction from free allicin (30.0 Ų) to Al₁₂N₁₂ complex (0.676 Ų) indicates markedly tighter atomic fluctuation at the binding interface.

Figure 1D. Experimental validation layer. (a) Half-life predictions vs. experimental literature across six temperature/pH conditions: R² = 0.9997, MAPE = 10.3% ^[27,28]. (b) IR vibrational frequency comparison: DFT-predicted vs. experimental wavenumbers for S=O stretch, S-S stretch, O-B stretch, and O-Al stretch; error labels indicate percentage deviation (maximum: +1.18% for O-Al) ^[1,7].

Figure 1E. Docking affinities and stability landscape. (a) Molecular docking binding energies (EDc, kcal/mol) for all four species against HER2, TNF-α, COVID-19 Mpro, and S. aureus — Allicin/Al₁₂N₁₂ consistently achieves the strongest affinity (most negative EDc) across all targets ^[1]. (b) Free allicin stability landscape across temperature (°C) and pH space; color scale = predicted t½ (days, capped at 300). ★ = physiological conditions (37 °C, pH 7.4); ▲ = standard lab storage (23 °C, pH 7.0); ■ = cold storage (4 °C, pH 7.0).

CONCLUSION

 This study presents the first Digital Twin framework for modeling aqueous degradation of a natural bioactive compound — allicin — and its Al₁₂N₁₂, B₁₂N₁₂, and C₂₄ nanocomplexes. The three-layer architecture successfully integrates DFT thermodynamic data [1], Gaussian Process MD surrogate emulation, and multi-source experimental validation into a unified, adaptive computational system.

Principal findings: (i) Allicin/Al₁₂N₁₂ achieves NPF = 5.5 × 10⁶ at physiological conditions (37 °C, pH 7.4), extending t½ from 3.6 days to beyond 9,999 days — resolving allicin's principal pharmacological limitation if confirmed experimentally. (ii) The GP surrogate accurately captures structural encapsulation signatures: SASA decreases from 13.26 nm² (free) to 6.96 nm² (Al₁₂N₁₂), and B-factor from 30.0 to 0.676 Ų. (iii) Stability prediction achieves R² = 0.9997, MAPE = 10.3%; IR frequencies within 1.18%. (iv) Allicin/Al₁₂N₁₂ ranks first in binding affinity against all four targets (HER2, TNF-α, COVID-19 Mpro, S. aureus) [1].

The data assimilation module allows recalibration in minutes as new experimentais datas (GROMACS trajectories, UV-vis, or stability measurements) become available. Future work will focus on: (a) validation with real GROMACS MD trajectories; (b) drug release kinetics modeling in simulated gastrointestinal fluid; (c) ADMET integration; and (d) extension to DADS and ajoene nanocomplexes.

CONFLICT OF INTEREST

       The authors declare no conflict of interest.

REFERENCES       

1. Mozafari ES, Baei MT, Tazikeh Lemeski E. Computational study of the therapeutic properties of allicin and its nanocomplexes using DFT and molecular docking techniques. Sci Rep. 2025; 15:23034. doi:10.1038/s41598-025-03293-0

2. Batiha GE, Beshbishy AM, Wasef LG, Elewa YHA, Al-Sagan AA, Abd El-Hack ME, et al. Chemical constituents and pharmacological activities of garlic (Allium sativum L.): A review. Nutrients. 2020;12(3):872. doi:10.3390/nu12030872

3. Stępień AE, Trojniak J, Tabarkiewicz J. Anti-cancer and anti-inflammatory properties of black garlic. Int J Mol Sci. 2024;25(3):1801. doi:10.3390/ijms25031801

4. Ankri S, Mirelman D. Antimicrobial properties of allicin from garlic. Microbes Infect. 1999;2(2):125–129. doi:10.1016/S1286-4579(99)80003-3

5. Cavallito CJ, Bailey JH. Allicin, the antibacterial principle of Allium sativum. I. Isolation, physical properties and antibacterial action. J Am Chem Soc. 1944;66(11):1950–1951. doi:10.1021/ja01239a048

6. Kodera Y, Ushijima M, Amano H, Suzuki J, Matsutomo T. Chemical and biological properties of S-1-propenyl-L-cysteine in aged garlic extract. Molecules. 2017;22(4):570. doi:10.3390/molecules22040570

7. Block E. The organosulfur chemistry of the genus Allium — implications for the organic chemistry of sulfur. Angew Chem Int Ed. 1992;31(9):1135–1178.

8. Li X, Chen X, Zhao Y, Wang H, Liu Z. Allicin inhibits the growth of HONE-1 and HNE1 human nasopharyngeal carcinoma cells by inducing ferroptosis. Neoplasma. 2024;71(3):243–254. doi:10.4149/neo_2024_240108n8

9. Catanzaro E, Canistro D, Pellicioni V, Vivarelli F, Fimognari C. Anticancer potential of allicin: A review. Pharmacol Res. 2022; 177:106118. doi:10.1016/j.phrs.2022.106118

10. Mousa AM, Yehia AH, El-Nekeety AA, Abdel-Wahhab BH, Hassan NS, Abdel-Wahhab MA. Could allicin alleviate trastuzumab-induced cardiotoxicity in a rat model through antioxidant, anti-inflammatory, and antihyperlipidemic properties? Life Sci. 2022; 302:120656. doi:10.1016/j.lfs.2022.120656

11. Choo S, Wong WT, Lim CJ, Bin Mohd Sani MH, Zulkefli ZN, Lim YY, et al. Review: antimicrobial properties of allicin used alone or in combination with other medications. Folia Microbiol. 2020;65(3):451–465. doi:10.1007/s12223-020-00786-5

12. Xu J, Chen X, Bai X, Ding C, Liu G, Sun Y, et al. Mechanism-aware digital twin for high-temperature creep prediction in Mo–Re alloys. Adv Sci. 2025. doi:10.1002/advs.202417547

13. Vairagade VS. Multi-scale theoretical modeling with molecular simulation framework for fly ash-based high-performance concrete. Sci Rep. 2025; 15:44895. doi:10.1038/s41598-025-28690-3

14. Haverkamp P, Anantheswar A, May M, Wadlig G, Hildebrandt J, Thiessat KA, et al. Digital Twin Road: value and implications involving data and application. Discov Appl Sci. 2025. doi:10.1007/s42452-025-06551-4

15. Savairam VD, Patil NA, Borate SR, Ghaisas MM, Shete RV. Allicin: A review of its important pharmacological activities. Pharmacol Res Mod Chin Med. 2023; 8:100283. doi:10.1016/j.prmcm.2023.100283

16. Ernzerhof M, Scuseria GE. Assessment of the Perdew-Burke-Ernzerhof exchange correlation functional. J Chem Phys. 1999; 110:5029–5036.

17. Becke AD, Johnson ER. A density-functional model of the dispersion interaction. J Chem Phys. 2005; 123:154101. doi:10.1063/1.2065267

18. Frisch MJ, Trucks GW, Schlegel HB, et al. Gaussian 09, Revision A.02. Wallingford CT: Gaussian Inc.; 2009.

19. Grimme S, Antony J, Ehrlich S, Krieg H. A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu. J Chem Phys. 2010; 132:154104. doi:10.1063/1.3382344

20. Bader RFW. Atoms in molecules. Acc Chem Res. 1985;18(1):9–15.

21. Bader RFW. A quantum theory of molecular structure and its applications. Chem Rev. 1991;91(5):893–928.

22. Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, et al. AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J Comput Chem. 2009;30(16):2785–2791.

23. Karelson M, Lobanov VS, Katritzky AR. Quantum-chemical descriptors in QSAR/QSPR studies. Chem Rev. 1996;96(3):1027–1044. doi:10.1021/cr950202r

24. Marenich AV, Cramer CJ, Truhlar DG. Universal solvation model based on solute electron density and on a continuum model of the solvent. J Phys Chem B. 2009;113(18):6378–6396. doi:10.1021/jp810292n

25. Swarna MR, Sathiyamoorthy T, Vijayaraj R, Ponmalar E, Mano S, Kumar AR. Understanding the adsorption performance of hetero-nanocages (C12–B6N6, C12–Al6N6, and B6N6–Al6N6) towards hydroxyurea anticancer drug: a comprehensive study using DFT. Nanoscale Adv. 2024. doi:10.1039/d4na00472h

26. Nair RGS, Nair AKN, Sun S. Adsorption of drugs on B12N12 and Al12N12 nanocages. RSC Adv. 2024;14(43):31756–31767. doi:10.1039/d4ra05586a

27. Miron T, Rabinkov A, Mirelman D, Wilchek M, Weiner L. The mode of action of allicin: Its ready permeability through phospholipid membranes may contribute to its biological activity. Biochim Biophys Acta. 2000;1463(1):20–30.

28. Lawson LD, Wang ZJ. Allicin and allicin-derived garlic compounds increase breath acetone through allyl methyl sulfide: Use in measuring allicin bioavailability. J Agric Food Chem. 2005;53(6):1974–1983.

29. Vaidya V, Ingold KU, Pratt DA. Garlic: Source of the ultimate antioxidants — sulfenic acids. Angew Chem Int Ed. 2009;48(1):157–160.

30. Marchese A, Barbieri R, Coppo E, Orhan IE, Daglia M, Nabavi SF, et al. Antifungal and antibacterial activities of allicin: A review. Trends Food Sci Technol. 2016; 52:49–56. doi:10.1016/j.tifs.2016.03.010